Gemini Code Assist enabled for PRs

[schweikert]

@auerswal and @gsnw-sebast FYI I enabled Google Code Assist in this repo. An example of what it does is visible here: #405.

I am not a huge fan of AI for code generation, but I do think that it can provide valuable feedback on pull requests. Let me know, if you find it more harmful than useful.

[schweikert]

I created here a PR with all changes since the last release that are committed in the develop branch:

#409

It has a couple of interesting comments from Google Code Assist. The "Highlights" section is also nice.

[auerswal]

I think that the first issue regarding an incorrect regular expression is an issue, but I did not verify it (yet). The second issue is correctly identifying an output formatting problem (and I'd assume that any existing test for this output would need to be adjusted, too). The third issue is wrong.

As they say, two out of three ain't bad. ;-)

I do not like the "Highlights", they are too imprecise.

The "AI code review" did find issues, but it also created additional work looking into a non-issue. IMHO this is OK, especially for new pull requests. All human reviewers need to understand that the AI does not understand anything, and question every bit of the AI's output.

[schweikert]

Yes, +1 on everything you said. Take the AI code review output always with a big grain of salt and keep in mind that it can be absolutely wrong (but often convincing).

[gsnw-sebast]

In the tests, gemini does not recognize the line break itself as a newline character. So there are a few false positives.
In general, however, I am open to AI verification because finding such small, non-critical errors works (see pull request #410).
There should be no blind trust.

[gsnw-sebast]

What do you think about the idea of using gemini code assist to create CHANGELOG.md or checking whether it could simplify this process? If that's even possible.